The Predictive Control Strategy For Wastewater Treatment Process Based On Activated Sludge Model

Activated sludge method is an effective method for the wastewater treatment and has been widely applied and studied. With the legislation on the strict wastewater discharge norm, it is necessary to adopt advanced modeling and control technology for the wastewater treatment process (WWTP) to improve water quality and reduce operating cost. But it is difficult to modeling and control for WWTP because of its strong feed disturbance and nonlinearity. Model predictive control (MPC) is widely used in the industrial process, so it is meaningful to study the model predictive control for the WWTP. In this paper a multi-model modeling method is studied and generalized predictive control strategy is used to optimize control for the Benchmark of the WWTP based on activated sludge model No1 (ASM1).In the paper clustering modeling method is used to establish multiple linear models expression of the WWTP, then multivariable generalized predictive control is used to track the setpoint change with constant feed and setpoint control with the feed fluctuations; considered the uncertainty of the WWTP, adaptive model correction is used online for the all sub-models and special single sub-model respectively to improve the model predictive accuracy. Because the clustering modeling is an unsupervised modeling method, a supervised multi-model modeling method on the base of clustering modeling method is proposed to improve the modeling accuracy.In order to meet the discharge norm of the contamination concentration limit, zone generalized predictive control is adopted which transforms the conventional quadratic objective of the MPC into a linear programming with soft constraints; besides, considering the wastewater treatment plant's energy-saving targets, a multi-objective predictive control is formulated and finally hierarchical multi-objective optimization algorithm is applied for on-line optimization solution.With slow speed measurement of the key components of the process, a multi-rate identification algorithm is designed to identify the rapid-rate model parameter and obtain intersample output estimation from the multi-rate data; then the proposed method is used for the WWTP to get the fast single-rate model from the sludge circumfluence flow to the nitrate concentration. Finally generalized predictive control is used to control the WWTP in the case of noise free and with noise in the output sample.